CN113657028B - Online aerosol optical thickness prediction method based on multi-source information - Google Patents
Online aerosol optical thickness prediction method based on multi-source information Download PDFInfo
- Publication number
- CN113657028B CN113657028B CN202110898847.6A CN202110898847A CN113657028B CN 113657028 B CN113657028 B CN 113657028B CN 202110898847 A CN202110898847 A CN 202110898847A CN 113657028 B CN113657028 B CN 113657028B
- Authority
- CN
- China
- Prior art keywords
- wind speed
- output
- data
- humidity
- optical thickness
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000003287 optical effect Effects 0.000 title claims abstract description 74
- 239000000443 aerosol Substances 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000013507 mapping Methods 0.000 claims abstract description 47
- 239000012634 fragment Substances 0.000 claims abstract description 28
- 238000004364 calculation method Methods 0.000 claims abstract description 19
- 230000006870 function Effects 0.000 claims abstract description 15
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 11
- 238000000513 principal component analysis Methods 0.000 claims abstract description 11
- 238000000605 extraction Methods 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims description 67
- 239000011159 matrix material Substances 0.000 claims description 51
- 230000003993 interaction Effects 0.000 claims description 19
- 238000012163 sequencing technique Methods 0.000 claims description 16
- 230000000694 effects Effects 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 10
- 238000009825 accumulation Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 230000001186 cumulative effect Effects 0.000 claims description 8
- 230000007613 environmental effect Effects 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 5
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 abstract description 2
- 230000008054 signal transmission Effects 0.000 abstract description 2
- 241000257303 Hymenoptera Species 0.000 description 9
- 230000008859 change Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000012937 correction Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 238000011835 investigation Methods 0.000 description 3
- 230000008033 biological extinction Effects 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000000053 physical method Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Mathematical Optimization (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Molecular Biology (AREA)
- Pure & Applied Mathematics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to an aerosol optical thickness online prediction method based on multi-source information, and belongs to the technical field of atmospheric channel optical signal transmission. The method comprises the steps of establishing a multi-source data acquisition system, carrying out feature extraction on atmospheric environment parameters, constructing a nonlinear mapping model based on a dynamic extreme learning machine, carrying out feature extraction through principal component analysis, and realizing fusion of features such as wind speed, temperature, humidity, sky background brightness and the like, wherein the output parameters of the model are aerosol optical thickness. The method has the advantages that the parameter estimation is carried out by using an artificial bee colony algorithm, so that the optimal selection of the weight and the bias is realized; according to the data fragments which arrive in real time, a nonlinear mapping model is updated, an online aerosol optical thickness prediction function is realized, a single hidden layer neural network based on a dynamic extreme learning machine is formed, the bottleneck of insufficient precision and difficult calculation caused by a physical model driving mode is broken through, and a solution is provided for constructing online estimation of optical parameters under various conditions.
Description
Technical Field
The invention belongs to the technical field of atmospheric channel optical signal transmission, and particularly relates to an aerosol optical thickness online prediction method based on multi-source information.
Background
Optical parameter estimation is a key element of optical information transmission, and in an atmospheric channel, due to the influence of turbulence, spot offset and angle of arrival fluctuation are often caused, so that difficulty is brought to optical parameter estimation. The optical thickness of the aerosol is an important support for judging the optical parameters, and has important significance for determining the values and change rules of the optical parameters. The model establishment and inversion calculation are carried out by a physical method, which can cause the bottleneck that modeling is difficult or can not be solved, influence the estimation precision of the optical parameters, and have great significance on accurately estimating the optical parameters and designing the optical communication system by carrying out online prediction on the optical thickness (Aerosol Optical Depth, AOD) of the aerosol in a data driving mode.
The Chinese patent publication No. 106446307A of AOD vertical correction effect evaluation method and System based on aerosol foundation data provides an aerosol data calculation method which comprises the steps of estimating an aerosol extinction coefficient near the ground and inverting the aerosol extinction coefficient based on visibility meter observation data, and vertically correcting the AOD by using a simulation method, wherein the influence of different weather on the aerosol thickness is not considered, and the change trend of the AOD cannot be predicted.
Chinese patent publication No. 106407634A, a method and a system for fusing satellite AOD data based on inverse variance weighted average, are provided. The invention can improve the coverage rate of satellite aerosol thickness data and the fusion precision requirement. Chinese patent publication No. 109213964A, satellite AOD product correction method integrating multisource characteristic geographic parameters, discloses an aerosol correction method, corrects satellite measured parameters through random forests and correction thresholds, and provides a reference for researching the effective concentration of PM2.5 in the atmosphere. However, these two methods are difficult to meet the condition of online change, and have strong dependence on the selection of the threshold value.
The Chinese patent publication No. 106096246A of the aerosol optical thickness estimation method based on PM2.5 and PM10 provides an aerosol estimation method, which is estimated by a universal gravitation neural network model, solves the problems that the inversion accuracy is not high and the real-time acquisition is difficult to realize, but the method does not consider the relation between the change of measured data and the update of the model, and cannot meet the requirement of automatic adjustment of the model. The Chinese patent publication No. 110186823A discloses an aerosol optical thickness inversion method, a bidirectional reflectance distribution function parameter database is established through historical data, and a corresponding aerosol optical thickness value is obtained through a table lookup method, but the method cannot calculate the generalization performance of a model, and a calculation result is only effective under specific assumption conditions.
Aiming at different requirements, the modeling mode of the extreme learning machine is widely applied, for example, china patent publication No. 109615003A of an ELM-CHMM-based power failure prediction method, china patent publication No. 109818798A of a KPCA and ELM-fused wireless sensor network intrusion detection system and method, and the extreme learning machine effectively supports a software function module of a complex system due to the characteristic of high convergence speed and clear analysis expression.
The online aerosol optical thickness prediction method based on the multi-source information is researched, so that not only can the optical parameters be effectively estimated, but also the design of an atmospheric optical communication system can be promoted, and the method has practical application value.
Disclosure of Invention
The invention provides an aerosol optical thickness online prediction method based on multi-source information, which fully considers the multi-source data information and the change trend of the data and provides a solution for online estimation of optical parameters under various conditions.
The technical scheme adopted by the invention is that the method comprises the following steps:
(1) Establishing a multi-source information acquisition system for acquiring and storing atmospheric environment parameters and optical parameters;
(2) Extracting features of atmospheric environment parameters to obtain wind speed features WS D Temperature characteristics TE D Humidity characteristics HU D Sky background brightness feature LU D An input parameter x is formed, and the optical thickness of the aerosol is an output parameter y;
(3) By adopting a data driving thought, a nonlinear mapping model based on a dynamic Extreme Learning Machine (ELM) is established, the relation between input parameters and output parameters is determined, and an artificial bee colony algorithm is utilized for optimization, so that the optimal solution of the input weight and bias of the nonlinear mapping model is determined;
(4) And (3) online updating the nonlinear mapping model, calculating errors of real values and predicted values of the real-time arrival data segments, quantitatively updating the nonlinear mapping model according to the errors, analyzing and calculating output weights of the nonlinear mapping model to obtain the aerosol optical thickness predicted value of the next data segment, realizing online aerosol optical thickness prediction, and fusing multi-source features through feature extraction, the nonlinear model and model updating strategies to realize online aerosol optical thickness prediction.
The multi-source information acquisition system in the step (1) comprises data acquisition equipment, an information interaction terminal and a database, wherein the data acquisition equipment comprises a laser radar, a ground weather station, a background radiometer and a corresponding data interface, the database stores data by adopting an SQL Server, the information interaction terminal performs information interaction with the data acquisition equipment, and acquired data information is stored in the database; the ground weather station, the background radiometer and the laser radar are used for realizing acquisition of atmospheric environmental parameters and optical parameters, wherein the atmospheric environmental parameters comprise atmospheric data wind speed WS, temperature TE, humidity HU provided by the ground weather station and sky background brightness LU acquired by the background radiometer, and the optical parameters are aerosol optical thickness and are acquired by the laser radar.
The atmospheric environment parameter characteristic extraction method in the step (2) is as follows:
performing principal component analysis on the atmospheric data stored in the data acquisition system, wherein the principal component analysis comprises wind speed WS, temperature TE, humidity HU and sky background brightness LU, and considering the accumulation effect of M days, a K-th day wind speed sample segment WS K Written in a matrix form as follows:
wherein ,wsKM Representing the wind speed value on the M th day in the K-th sample segment, for the wind speed sample matrix segment WS K The wind speed value of the wind turbine is normalized:
wherein ,ws′km Represents the normalized value of wind speed, mu, on the m-th day in the kth sample segment m 、σ m Respectively represent the mth day windThe mean value and variance of the fast sample fragments are processed to obtain a wind speed standardization matrix WS' K And calculates a wind speed characteristic value lambda WS,1 ,λ WS,2 ,…λ WS,M Corresponding wind speed eigenvector eta WS,1 ,η WS,2 ,…η WS,M Sequencing the wind speed characteristic values from large to small, calculating the contribution degree of the wind speed characteristic values, and when the contribution degree of the selected wind speed characteristic values is larger than a threshold value theta 1 In this case, the number of principal components Num is determined 1 ；
The corresponding wind speed characteristic vector set is the wind speed characteristic
Considering the cumulative effect on day M, for the temperature sample fragment TE on day K K Carrying out standardization treatment to obtain a temperature standardization matrix TE' K Calculating a temperature characteristic value lambda TE,1 ,λ TE,2 ,…λ TE,M Corresponding temperature characteristic vector eta TE,1 ,η TE,2 ,…η TE,M Sequencing the temperature characteristic values from large to small, calculating the contribution degree of the temperature characteristic values, and when the contribution degree of the selected temperature characteristic values is larger than a threshold value theta 2 In this case, the number of principal components Num is determined 2 ；
The corresponding temperature characteristic vector set is the temperature characteristic
Considering the cumulative effect on day M, for day K humidity sample fragment HU K Performing standardization treatment to obtain a humidity standardization matrix HU' K Calculating humidity characteristic value lambda HU,1 ,λ HU,2 ,…λ HU,M Corresponding humidity characteristic vector eta HU,1 ,η HU,2 ,…η HU,M Sequencing the humidity characteristic values from large to small, calculating the contribution degree of the humidity characteristic values, and when the contribution degree of the selected humidity characteristic values is larger than a threshold value theta 3 In this case, the number of principal components Num is determined 3
The corresponding humidity characteristic vector set is the humidity characteristic
Taking the accumulation effect of M days into consideration, for the K-th sky background brightness sample segment LU K Performing standardization processing to obtain a sky background brightness standardization matrix LU' K Calculating sky background brightness characteristic value lambda LU,1 ,λ LU,2 ,…λ LU,M Corresponding sky background brightness characteristic vector eta LU,1 ,η LU,2 ,…η LU,M Sequencing the sky background brightness characteristic values from large to small, calculating contribution degrees of the sky background brightness characteristic values, and when the contribution degrees of the selected sky background brightness characteristic values are larger than a threshold value theta 4 In this case, the number of principal components Num is determined 4 ；
The corresponding sky background brightness characteristic vector set is the sky background brightness characteristic
The input parameter x is the wind speed characteristic TE D Temperature characteristics TE D Humidity characteristics HU D Sky background brightness feature LU D Is x= (WS) D ,TE D ,HU D ,LU D )。
In the step (3) of the inventionThe construction of the nonlinear mapping model comprises a 3-layer neural network structure of an input layer, a hidden layer and an output layer, the newly arrived sample data is considered, the data sampling length is N, the input layer is N nodes in the initial stage, the hidden layer is L nodes, the output layer is 1 node, and the input parameter x and the output parameter y are sampled and combined to obtain the sample datax i The ith input parameter representing x, y i The i output parameter representing y, the nonlinear mapping model input-output relationship is as follows:
wherein a and b are input weight and bias respectively, beta is output layer weight, L is hidden layer node number of the neural network,in order to activate the function, the mapping relation among the nodes is considered on the whole, the weight and the bias between the nodes of the input layer and the hidden layer are randomly selected according to the extreme learning machine theory, the output weight beta is the only variable to be solved in the whole network, and the formula (7) can be expressed as the following matrix form through Moore-Penrose generalized inverse calculation:
y＝Hβ······························(8)
wherein y= [ y ] 1 ,y 2 ,...,y N ] T The aerosol optical thickness of the current segment is represented, a hidden layer matrix H in a nonlinear model is introduced for calculation, H is a random matrix of N rows and L columns, and the specific expression is:
wherein ,for activating the function, realizing the nonlinear mapping, ensuring that the derivative exists in the interval of the function, and for further calculating the output weight beta, calculating Moore-Penrose generalized inverse of H>
H + ＝(H T H) -1 H T ·················(10)
Obtaining an optimal solution beta of the output weight beta by adopting a form of solving Moore-Penrose generalized inverse * ：
β * ＝H + y····························(11)
Further optimizing input weights and offsets, and performing character string coding on the weights and offsets to obtain a candidate solution vector set E= (a, b), wherein the specific form is as follows:
the solution vector with the boundaries of the candidate vector set being the maximum and minimum, denoted here as E max = { max (E) |e E } and E min = { min (E) |e E }, where E represents a set of sets where each element satisfies a boundary condition, optimization is performed according to an artificial bee colony algorithm, and under an initial condition, a solution vector of input weight and bias is expressed as:
E h,j ＝E min,j +ω×(E max,j -E min,j )···················(13)
E min,j and Emax,j Respectively represent the maximum value and the minimum value of the j-th input solution vector, and omega is uniformly distributed in the interval [ -1,1 [ -1 ]]And carrying out optimal solution vector search to obtain the following updating strategy:
wherein ,Eh,j Representing the current solution vector of the current set of values,representing new feasible solution vectors, h, j and r are all indication marks, u h,j In the interval [ -1,1]Randomly changing in the range, calculating the fitness, and selecting the optimal solution vector:
wherein num represents the iteration number, and after iteration is completed, the optimal solution vector is updated to obtain the optimal input weight a * Bias b * Then, the expression optimal solution of the nonlinear model is:
the nonlinear mapping model updating method in the step (4) is as follows:
in the initial stage, the output weight is obtainedNew sample fragment->After arrival, the errors of the true value and the predicted value of the real-time arrival data fragment are calculated and expressed as:
wherein ,N0 and N1 Respectively representing the starting position and the stopping position of the current sample fragment, H 1 and y1 Hidden layer output and tag set, beta, representing newly arrived segments, respectively 1 Representing the current nonlinear model output weight, and obtaining beta based on the current output weight 1 The expression is:
considering the above expression in two parts, an intermediate variable is introducedAccording to the matrix multiplication combining law, the expression of the former part of matrix is:
in the current fragment state, H 1 and K0 Are known conditions and are further substituted into the integral output layer weight expression, so that the output weight beta of the model updating stage can be obtained 1 Is represented by the expression:
further performing iterative calculation to obtain an intermediate variable K corresponding to the (k+1) th segment k+1 Expression type
Further performing iterative computation to obtain the nonlinear mapping model output weight beta corresponding to the (k+1) th segment k+1 Is the optimal solution of (a):
the model expression after online updating is as follows:
substituting the input parameter x into the updated model, the aerosol optical thickness y can be calculated, and online prediction of the aerosol optical thickness is realized.
The invention establishes a multi-source data acquisition system and realizes the information interaction from the data acquisition equipment to the database through the information interaction terminal. The data acquisition equipment is used for acquiring data such as wind speed, temperature, humidity and sky background brightness, transmitting the data to the information interaction terminal through the data interface, and realizing data storage through data processing operation. On the basis of a multi-source data acquisition system, a nonlinear mapping model is constructed, a neural network structure is designed, and optimization of the weight and bias of the neural network is realized, so that a mapping model of the aerosol optical thickness is obtained. And the nonlinear model is updated through the data acquired in real time, so that the online prediction method for the optical thickness of the aerosol is realized.
The invention has the advantages that:
(1) According to the invention, a multisource data acquisition system is considered, the environmental parameters and the optical parameters are stored, a nonlinear mapping model is established through the data driving thought, the online prediction of the aerosol optical parameters is realized, and the bottlenecks of insufficient precision and difficulty in calculation caused by a physical model driving mode are broken through.
(2) Feature extraction is performed through principal component analysis, a neural network is built by utilizing a dynamic extreme learning mechanism, weight and bias are optimized, and input and output relations adaptable to an online environment are obtained.
(3) By means of the multisource information processing scheme, the model is quantitatively adjusted according to the current data fragment, re-learning of the model is avoided, automation of the whole prediction process is achieved, a solution is provided for intelligent processing of a complex system, and the method has certain universality for engineering practice.
Drawings
FIG. 1 is a multi-source data acquisition system of the present invention;
FIG. 2 is a non-linear mapping model of the present invention;
fig. 3 is a flow chart of the present invention.
Detailed Description
Comprises the following steps:
(1) Establishing a multi-source information acquisition system for acquiring and storing atmospheric environment parameters and optical parameters;
(2) Extracting features of atmospheric environment parameters to obtain wind speed features WS D Temperature characteristics TE D Humidity characteristics HU D Sky background brightness feature LU D An input parameter x is formed, and the optical thickness of the aerosol is an output parameter y;
(3) By adopting a data driving thought, a nonlinear mapping model based on a dynamic Extreme Learning Machine (ELM) is established, the relation between input parameters and output parameters is determined, and an artificial bee colony algorithm is utilized for optimization, so that the optimal solution of the input weight and bias of the nonlinear mapping model is determined;
(4) And (3) online updating the nonlinear mapping model, calculating errors of real values and predicted values of the real-time arrival data segments, quantitatively updating the nonlinear mapping model according to the errors, analyzing and calculating output weights of the nonlinear mapping model to obtain the aerosol optical thickness predicted value of the next data segment, realizing online aerosol optical thickness prediction, and fusing multi-source features through feature extraction, the nonlinear model and model updating strategies to realize online aerosol optical thickness prediction.
The multi-source information acquisition system in the step (1) comprises data acquisition equipment, an information interaction terminal and a database, wherein the data acquisition equipment comprises a laser radar, a ground weather station, a background radiometer and a corresponding data interface, the database stores data by adopting an SQL Server, the information interaction terminal performs information interaction with the data acquisition equipment, and acquired data information is stored in the database; the ground weather station, the background radiometer and the laser radar are used for realizing acquisition of atmospheric environmental parameters and optical parameters, wherein the atmospheric environmental parameters comprise atmospheric data wind speed WS, temperature TE, humidity HU provided by the ground weather station and sky background brightness LU acquired by the background radiometer, and the optical parameters are aerosol optical thickness and are acquired by the laser radar;
the atmospheric environment parameter characteristic extraction method in the step (2) is as follows:
performing principal component analysis on the atmospheric data stored in the data acquisition system, wherein the principal component analysis comprises wind speed WS, temperature TE, humidity HU and sky background brightness LU, and considering the accumulation effect of M days, a K-th day wind speed sample segment WS K Written in a matrix form as follows:
wherein ,wsKM Representing the wind speed value on the M th day in the K-th sample segment, for the wind speed sample matrix segment WS K The wind speed value of the wind turbine is normalized:
wherein ,ws′km Represents the normalized value of wind speed, mu, on the m-th day in the kth sample segment m 、σ m Respectively representing the mean value and the variance of the sample fragments of the wind speed on the m th day, and obtaining a wind speed standardization matrix WS 'after processing' K And calculates a wind speed characteristic value lambda WS,1 ,λ WS，2 ,…λ WS,M Corresponding wind speed eigenvector eta WS,1 ,η WS,2 ,…η WS,M Sequencing the wind speed characteristic values from large to small, calculating the contribution degree of the wind speed characteristic values, and when the contribution degree of the selected wind speed characteristic values is larger than a threshold value theta 1 In this case, the number of principal components Num is determined 1 ；
The corresponding wind speed characteristic vector set is the wind speed characteristic
Considering the cumulative effect on day M, for the temperature sample fragment TE on day K K Performing standardization treatment to obtain temperature standardChemical matrix TE' K Calculating a temperature characteristic value lambda TE,1 ,λ TE,2 ,…λ TE,M Corresponding temperature characteristic vector eta TE,1 ,η TE,2 ,…η TE,M Sequencing the temperature characteristic values from large to small, calculating the contribution degree of the temperature characteristic values, and when the contribution degree of the selected temperature characteristic values is larger than a threshold value theta 2 In this case, the number of principal components Num is determined 2 ；
The corresponding temperature characteristic vector set is the temperature characteristic
Considering the cumulative effect on day M, for day K humidity sample fragment HU K Performing standardization treatment to obtain a humidity standardization matrix HU' K Calculating humidity characteristic value lambda HU,1 ,λ HU,2 ,…λ HU,M Corresponding humidity characteristic vector eta HU,1 ,η HU,2 ,…η HU,M Sequencing the humidity characteristic values from large to small, calculating the contribution degree of the humidity characteristic values, and when the contribution degree of the selected humidity characteristic values is larger than a threshold value theta 3 In this case, the number of principal components Num is determined 3
The corresponding humidity characteristic vector set is the humidity characteristic
Taking the accumulation effect of M days into consideration, for the K-th sky background brightness sample segment LU K Performing standardization processing to obtain a sky background brightness standardization matrix LU' K Calculating sky background brightness characteristic value lambda LU,1 ,λ LU,2 ,…λ LU,M Corresponding sky background brightness characteristic directionQuantity eta LU,1 ,η LU,2 ,…η LU,M Sequencing the sky background brightness characteristic values from large to small, calculating contribution degrees of the sky background brightness characteristic values, and when the contribution degrees of the selected sky background brightness characteristic values are larger than a threshold value theta 4 In this case, the number of principal components Num is determined 4 ；
The corresponding sky background brightness characteristic vector set is the sky background brightness characteristic
The input parameter x is the wind speed characteristic TE D Temperature characteristics TE D Humidity characteristics HU D Sky background brightness feature LU D Is x= (WS) D ,TE D ,HU D ,LU D )；
The nonlinear mapping model in the step (3) is constructed by a 3-layer neural network structure comprising an input layer, a hidden layer and an output layer, the newly arrived sample data is considered, the data sampling length is N, the input layer is N nodes in the initial stage, the hidden layer is L nodes, the output layer is 1 node, and the input parameter x and the output parameter y are sampled and combined to obtain the sample datax i The ith input parameter representing x, y i The i output parameter representing y, the nonlinear mapping model input-output relationship is as follows:
wherein a and b are input weight and bias respectively, beta is output layer weight, L is hidden layer node number of the neural network,in order to activate the function, the mapping relation among the nodes is considered on the whole, the weight and the bias between the nodes of the input layer and the hidden layer are randomly selected according to the extreme learning machine theory, the output weight beta is the only variable to be solved in the whole network, and the formula (7) can be expressed as the following matrix form through Moore-Penrose generalized inverse calculation:
y＝Hβ·····························(8)
wherein y= [ y ] 1 ,y 2 ,...,y N ] T The aerosol optical thickness of the current segment is represented, a hidden layer matrix H in a nonlinear model is introduced for calculation, H is a random matrix of N rows and L columns, and the specific expression is:
wherein ,for activating the function, realizing the nonlinear mapping, ensuring that the derivative exists in the interval of the function, and for further calculating the output weight beta, calculating Moore-Penrose generalized inverse of H>
H + ＝(H T H) -1 H T ·················(10)
Obtaining an optimal solution beta of the output weight beta by adopting a form of solving Moore-Penrose generalized inverse * ：
β * ＝H + y····························(11)
Further optimizing input weights and offsets, and performing character string coding on the weights and offsets to obtain a candidate solution vector set E= (a, b), wherein the specific form is as follows:
the solution vector with the boundaries of the candidate vector set being the maximum and minimum, denoted here as E max = { max (E) |e E } and E min = { min (E) |e E }, where E represents a set of sets where each element satisfies a boundary condition, optimization is performed according to an artificial bee colony algorithm, and under an initial condition, a solution vector of input weight and bias is expressed as:
E h,j ＝E min,j +ω×(E max,j -E min,j )···················(13)
E min,j and Emax,j Respectively represent the maximum value and the minimum value of the j-th input solution vector, and omega is uniformly distributed in the interval [ -1,1 [ -1 ]]And carrying out optimal solution vector search to obtain the following updating strategy:
wherein ,Eh,j Representing the current solution vector of the current set of values,and representing a new feasible solution vector, wherein h, j and r are all indication marks. u (u) h,j In the interval [ -1,1]Random variations within the range. Calculating fitness and selecting an optimal solution vector:
wherein num represents the iteration number, and after iteration is completed, the optimal solution vector is updated to obtain the optimal input weight a * Bias b * Then, the expression optimal solution of the nonlinear model is:
the nonlinear mapping model updating method in the step (4) is as follows:
in the initial stage, the output weight is obtainedNew sample fragment->After arrival, the errors of the true value and the predicted value of the real-time arrival data fragment are calculated and expressed as:
wherein ,N0 and N1 Respectively representing the starting position and the stopping position of the current sample fragment, H 1 and y1 Hidden layer output and tag set, beta, representing newly arrived segments, respectively 1 Representing the current nonlinear model output weight, and obtaining beta based on the current output weight 1 The expression is:
considering the above expression in two parts, an intermediate variable is introducedAccording to the matrix multiplication combining law, the expression of the former part of matrix is:
in the current fragment state, H 1 and K0 Are known conditions and are further substituted into the integral output layer weight expression, so that the output weight beta of the model updating stage can be obtained 1 Is represented by the expression:
further performing iterative calculation to obtain an intermediate variable K corresponding to the (k+1) th segment k+1 Expression type
Further performing iterative computation to obtain the nonlinear mapping model output weight beta corresponding to the (k+1) th segment k+1 Is the optimal solution of (a):
the model expression after online updating is as follows:
substituting the input parameter x into the updated model, the aerosol optical thickness y can be calculated, and online prediction of the aerosol optical thickness is realized. In the whole solving process, the nonlinear model does not need to be relearned, and model weight can be updated according to the arrived data, so that the online prediction function of the aerosol optical thickness is realized.
The present invention will be described in detail with reference to the accompanying drawings.
(1) And constructing a multisource data acquisition system, wherein the specific form is shown in figure 1, a laser radar acquires aerosol optical thickness information, a ground weather station acquires wind speed, temperature and humidity information, a background radiometer acquires sky background brightness information, and an information interaction terminal is a common computer and can realize information interaction with data acquisition equipment. The SQLServer software is loaded in the database, and the data storage function is realized by maintaining the data table;
the laser radar is connected with the information interaction terminal through an RJ45 network interface, the background radiometer and the ground weather station adopt an RS232 interface to carry out serial port communication with the information interaction terminal, the information interaction terminal sends an acquisition command to the data acquisition equipment in the process of information acquisition, the data acquisition equipment transmits data information to the information interaction terminal through a data interface to realize information interaction,
(2) And realizing feature extraction. The collected data are arranged, the accumulation effects of wind speed, temperature, humidity and sky background brightness are respectively considered through principal component analysis, 1 day is used as a data collection period, each half hour interval is 1 sampling point, each day is 48 sample vectors, and the data of 7 days before the current time node are considered to form a sample segment matrix of 48 rows and 7 columns;
performing principal component analysis on the atmospheric data stored in the data acquisition system, wherein the principal component analysis comprises wind speed WS, temperature TE, humidity HU and sky background brightness LU, and considering the accumulation effect of M days, a K-th day wind speed sample segment WS K Can be written in a matrix form as follows:
wherein ,wsKM Representing the wind speed value on the M th day in the K-th sample segment, for the wind speed sample matrix segment WS K The wind speed value of the wind turbine is normalized:
wherein ,ws′km Represents the normalized value of wind speed, mu, on the m-th day in the kth sample segment m 、σ m Respectively representing the mean value and the variance of the sample fragments of the wind speed on the m th day, and obtaining a wind speed standardization matrix WS 'after processing' K And calculates a wind speed characteristic value lambda WS,1 ,λ WS,2 ,…λ WS,M Corresponding wind speed eigenvector eta WS,1 ,η WS,2 ,…η WS,M Sequencing the wind speed characteristic values from large to small, calculating the contribution degree of the wind speed characteristic values, and when the wind speed characteristic values are selectedContribution degree is greater than threshold value theta 1 When=0.90, the number Num of principal components is determined 1 ；
The corresponding wind speed characteristic vector set is the wind speed characteristic
Considering the cumulative effect on day M, for the temperature sample fragment TE on day K K Carrying out standardization treatment to obtain a temperature standardization matrix TE' K Calculating a temperature characteristic value lambda TE,1 ,λ TE,2 ,…λ TE,M Corresponding temperature characteristic vector eta TE,1 ,η TE,2 ,…η TE,M Sequencing the temperature characteristic values from large to small, calculating the contribution degree of the temperature characteristic values, and when the contribution degree of the selected temperature characteristic values is larger than a threshold value theta 2 When=0.85, the number of principal components Num is determined 2
The corresponding temperature characteristic vector set is the temperature characteristic
Considering the cumulative effect on day M, for day K humidity sample fragment HU K Performing standardization treatment to obtain a humidity standardization matrix HU' K Calculating humidity characteristic value lambda HU,1 ,λ HU,2 ,…λ HU,M Corresponding humidity characteristic vector eta HU,1 ,η HU,2 ,…η HU,M Sequencing the humidity characteristic values from large to small, calculating the contribution degree of the humidity characteristic values, and when the contribution degree of the selected humidity characteristic values is larger than a threshold value theta 3 When=0.85, the number of principal components Num is determined 3
The corresponding humidity characteristic vector set is the humidity characteristic
Taking the accumulation effect of M days into consideration, for the K-th sky background brightness sample segment LU K Performing standardization processing to obtain a sky background brightness standardization matrix LU' K Calculating sky background brightness characteristic value lambda LU,1 ,λ LU,2 ,…λ LU,M Corresponding sky background brightness characteristic vector eta LU,1 ,η LU,2 ,…η LU,M Sequencing the sky background brightness characteristic values from large to small, calculating contribution degrees of the sky background brightness characteristic values, and when the contribution degrees of the selected sky background brightness characteristic values are larger than a threshold value theta 4 When=0.80, the number of principal components Num is determined 4 ；
The corresponding sky background brightness characteristic vector set is the sky background brightness characteristic
The input parameter x is the wind speed characteristic TE D Temperature characteristics TE D Humidity characteristics HU D Sky background brightness feature LU D Is x= (WS) D ,TE D ,HU D ,LU D )；
(3) By adopting a data driving thought, a nonlinear mapping model based on a dynamic Extreme Learning Machine (ELM) is established, the relation between input parameters and output parameters is determined, and an artificial bee colony algorithm is utilized for optimization, so that the optimal solution of the input weight and bias of the nonlinear mapping model is determined; wherein:
construction of nonlinear mapping model comprising input layer, hidden layer, and output layer of 3-layer godThrough a network structure, newly arrived sample data is considered, the data sampling length is 48, the input layer is 48 nodes, the hidden layer is 50 nodes, the output layer is 1 node, wherein the input layer nodes and the hidden layer are in full correlation connection, the hidden layer and the output layer are in full correlation connection, the input layer weight is a matrix of 48 rows and 50 columns, the offset is a matrix of 1 row and 50 columns, and the input parameter x and the output parameter y are sampled and combined to obtain the sample datax i The ith input parameter representing x, y i The i output parameter representing y, the nonlinear mapping model input-output relationship is as follows:
wherein a and b are input weight and bias respectively, beta is output layer weight, L is hidden layer node number of the neural network,in order to activate the function, the mapping relation among the nodes is considered on the whole, the weight and the bias between the nodes of the input layer and the hidden layer are randomly selected according to the extreme learning machine theory, the output weight beta is the only variable to be solved in the whole network, and the formula (7) can be expressed as the following matrix form through Moore-Penrose generalized inverse calculation:
y＝Hβ·····················(8)
wherein y= [ y ] 1 ,y 2 ,...,y N ] T The aerosol optical thickness of the current segment is represented, a hidden layer matrix H in a nonlinear model is introduced for calculation, H is a random matrix of 48 rows and 50 columns, and the specific expression is:
wherein the nonlinear modeThe input parameter may be expressed as x= [ x 1 ,x 2 ,...,x N ] T A plurality of features are fused;
output weights are calculated. The weight and bias between the input layer node and the hidden layer are randomly selected, and the weight between the hidden layer and the output layer is calculated through M-P generalized inverse. Obtaining output weight through M-P generalized inverse, generalized inverse of hidden layer matrix H
H + ＝(H T H) -1 H T ·········(10)
Satisfying the condition of minimum error, beta is the only variable to be solved in the whole network, y= [ y ] 1 ,y 2 ,...,y N ] T Representing the aerosol optical thickness value of the current segment. Obtaining an expression of the output weight through M-P generalized inverse;
β * ＝H + y····························(11)
further optimizing the weight and the bias, and carrying out character string coding on the weight and the bias to obtain a candidate vector set, wherein the specific form is as follows:
the candidate vector sets are bounded by maximum and minimum solution vectors, E represents a set of sets for which each element satisfies a boundary condition, E max = { max (E) |e E } and E min = { min (E) |e E }. The method is characterized in that the method is optimized according to an artificial bee colony algorithm, food searching is carried out by analog bees of the artificial bee colony algorithm, multi-parameter optimization is achieved, three forms are mainly considered, and the information searching task is completed by employing bees, following bees and investigation bees. In the process of carrying out optimal weight and bias solving, hiring bees to search information first, feeding back the searched result, receiving the fed back information by the bees, and continuing to surround the optimal vectorSearching, and if a better solution is found, updating; otherwise, giving up the current solution, and continuing searching by the investigation bees. Finally, hiring bees to realize local search nearby the optimal solution, finishing global search by the investigation bees, converging the whole algorithm, and obtaining the optimal solution vector of nonlinear mapping model weight and bias through optimization;
at the maximum and minimum E of the j-th input solution vector min,j and Emax,j On the basis of (1) calculating boundary combinations under initial conditions, the weight solution vector is expressed as:
E h,j ＝E min,j +ω×(E max，j -E min,j )··················(13)
determining update parameters, which in this case result in the interval [ -1,1]U randomly varying in range h,j Then, through optimal vector searching, a candidate vector updating strategy is obtained:
by a random variable u h,j The calculation of fitness under different weights and bias conditions is realized:
here, the iteration number num is 50, and after the iteration is completed, the optimal solution vector is updated to obtain the optimal input weight a * Bias b * And synthesizing weights and biases of all layers of the neural network to obtain a relation model of input parameters and output parameters:
the neural network input layer is 48 nodes, the hidden layer is 50 nodes, and the output layer is 1 node in the initial stage. Determining all parameters of the nonlinear mapping model to obtain the relationship between the input parameters and the output parametersNamely, the analytic relation of wind speed, temperature, humidity, sky background brightness and aerosol optical thickness, and the output weight obtained in the initial stage is expressed as:
(4) Model update, new sample fragmentAfter arrival, calculating the approximation error of the sample according to the acquired data fragment, wherein the objective function is expressed as:
N 0 and N1 Respectively representing the starting position and the stopping position of the current sample fragment, H 1 and Y1 Hidden layer output and tag set, beta, representing newly arrived segments, respectively 1 The table represents the current nonlinear model weights. Based on the current output weight, the obtained output layer weight expression is:
taking the front part and the back part of the expression into consideration respectively, introducing intermediate variablesThe expression of the previous part of matrix is calculated by matrix multiplication and combination law as follows: />
Consider beta 1 The latter part of the expression is expanded by using the matrix multiplication principle to obtain:
based on the current output weight, substituting new input parameters into a nonlinear model, obtaining an expression of the output weight through matrix transformation, and at the current moment, H 1 and K0 All are known conditions, and the output weight of the calculation model updating stage is as follows:
in the calculation process, an intermediate variable random matrix K is introduced to realize the mapping expression of the input parameters and the output parameters of the current segment, and the expression of the (k+1) th segment after each iteration is further obtained:
calculating the optical thickness of the aerosol, substituting the input parameter information of the new segment into a nonlinear model, and calculating through a neural network to obtain an output parameter value;
finally, through inverse normalization, the prediction result of the final aerosol optical thickness is obtained.
Claims (2)
1. An online aerosol optical thickness prediction method based on multi-source information is characterized by comprising the following steps:
(1) Establishing a multi-source information acquisition system for acquiring and storing atmospheric environment parameters and optical parameters;
(2) For atmospheric environment parametersExtracting row characteristics to obtain wind speed characteristics WS D Temperature characteristics TE D Humidity characteristics HU D Sky background brightness feature LU D An input parameter x is formed, and the optical thickness of the aerosol is an output parameter y;
the atmospheric environment parameter characteristic extraction method comprises the following steps:
performing principal component analysis on the atmospheric data stored in the data acquisition system, wherein the principal component analysis comprises wind speed WS, temperature TE, humidity HU and sky background brightness LU, and considering the accumulation effect of M days, a K-th day wind speed sample segment WS K Written in a matrix form as follows:
wherein ,wsKM Representing the wind speed value on the M th day in the K-th sample segment, for the wind speed sample matrix segment WS K The wind speed value of the wind turbine is normalized:
wherein ,ws′km Represents the normalized value of wind speed, mu, on the m-th day in the kth sample segment m 、σ m Respectively representing the mean value and the variance of the sample fragments of the wind speed on the m th day, and obtaining a wind speed standardization matrix WS 'after processing' K And calculates a wind speed characteristic value lambda WS,1 ,λ WS,2 ,...λ WS,M Corresponding wind speed eigenvector eta WS,1 ,η WS,2 ,...η WS,M Sequencing the wind speed characteristic values from large to small, calculating the contribution degree of the wind speed characteristic values, and when the contribution degree of the selected wind speed characteristic values is larger than a threshold value theta 1 In this case, the number of principal components Num is determined 1 ；
Corresponding wind speedThe feature vector set is the wind speed feature
Considering the cumulative effect on day M, for the temperature sample fragment TE on day K K Carrying out standardization treatment to obtain a temperature standardization matrix TE' K Calculating a temperature characteristic value lambda TE,1 ,λ TE,2 ,...λ TE,M Corresponding temperature characteristic vector eta TE,1 ,η TE,2 ,...η TE,M Sequencing the temperature characteristic values from large to small, calculating the contribution degree of the temperature characteristic values, and when the contribution degree of the selected temperature characteristic values is larger than a threshold value theta 2 In this case, the number of principal components Num is determined 2 ；
The corresponding temperature characteristic vector set is the temperature characteristic
Considering the cumulative effect on day M, for day K humidity sample fragment HU K Performing standardization treatment to obtain a humidity standardization matrix HU' K Calculating humidity characteristic value lambda HU,1 ,λ HU,2 ,...λ HU,M Corresponding humidity characteristic vector eta HU,1 ,η HU,2 ,...η HU,M Sequencing the humidity characteristic values from large to small, calculating the contribution degree of the humidity characteristic values, and when the contribution degree of the selected humidity characteristic values is larger than a threshold value theta 3 In this case, the number of principal components Num is determined 3
The corresponding humidity characteristic vector set is the humidity characteristic
Taking the accumulation effect of M days into consideration, for the K-th sky background brightness sample segment LU K Performing standardization processing to obtain a sky background brightness standardization matrix LU' K Calculating sky background brightness characteristic value lambda LU,1 ,λ LU,2 ,...λ LU,M Corresponding sky background brightness characteristic vector eta LU,1 ,η LU,2 ,...η LU,M Sequencing the sky background brightness characteristic values from large to small, calculating contribution degrees of the sky background brightness characteristic values, and when the contribution degrees of the selected sky background brightness characteristic values are larger than a threshold value theta 4 In this case, the number of principal components Num is determined 4 ；
The corresponding sky background brightness characteristic vector set is the sky background brightness characteristic
The input parameter x is the wind speed characteristic TE D Temperature characteristics TE D Humidity characteristics HU D Sky background brightness feature LU D Is x= (WS) D ,TE D ,HU D ,LU D )；
(3) By adopting a data driving thought, a nonlinear mapping model based on a dynamic Extreme Learning Machine (ELM) is established, the relation between input parameters and output parameters is determined, and an artificial bee colony algorithm is utilized for optimization, so that the optimal solution of the input weight and bias of the nonlinear mapping model is determined;
the construction of the nonlinear mapping model comprises a 3-layer neural network structure of an input layer, a hidden layer and an output layer, the newly arrived sample data is considered, the data sampling length is N, the input layer is N nodes, the hidden layer is L nodes, the output layer is 1 node, and the input parameter x and the output parameter y are sampled and combined to obtain sample datax i The ith input parameter representing x, y i The i output parameter representing y, the nonlinear mapping model input-output relationship is as follows:
wherein a and b are input weight and bias respectively, beta is output layer weight, L is hidden layer node number of the neural network,in order to activate the function, the mapping relation among the nodes is considered on the whole, the weight and the bias between the nodes of the input layer and the hidden layer are randomly selected according to the extreme learning machine theory, the output weight beta is the only variable to be solved in the whole network, and the formula (7) can be expressed as the following matrix form through Moore-Penrose generalized inverse calculation:
y＝Hβ································ (8)
wherein y= [ y ] 1 ,y 2 ,...,y N ] T The aerosol optical thickness of the current segment is represented, a hidden layer matrix H in a nonlinear model is introduced for calculation, H is a random matrix of N rows and L columns, and the specific expression is:
wherein ,for activating the function, realizing the nonlinear mapping, ensuring that the derivative exists in the interval of the function, and for further calculating the output weight beta, calculating Moore-Penrose generalized inverse of H>
H + ＝(H T H) -1 H T ···························(10)
Obtaining an optimal solution beta of the output weight beta by adopting a form of solving Moore-Penrose generalized inverse * ：
β * ＝H + y···························(11)
Further optimizing input weights and offsets, and performing character string coding on the weights and offsets to obtain a candidate solution vector set E= (a, b), wherein the specific form is as follows:
the solution vector with the boundaries of the candidate vector set being the maximum and minimum, denoted here as E max = { max (E) |e E } and E min = { min (E) |e E }, where E represents a set of sets where each element satisfies a boundary condition, optimization is performed according to an artificial bee colony algorithm, and under an initial condition, a solution vector of input weight and bias is expressed as:
E h,j ＝E min,j +ω×(E max,j -E min,j )····················(13)
E min,j and Emax,j Respectively represent the maximum value and the minimum value of the j-th input solution vector, and omega is uniformly distributed in the interval [ -1,1 [ -1 ]]And carrying out optimal solution vector search to obtain the following updating strategy:
wherein ,Eh,j Representing the current solution vector of the current set of values,representing a newFeasible solution vectors, h, j and r are all indication marks, u h,j In the interval [ -1,1]Randomly changing in the range, calculating the fitness, and selecting the optimal solution vector:
wherein num represents the iteration number, and after iteration is completed, the optimal solution vector is updated to obtain the optimal input weight a * Bias b * Then, the expression optimal solution of the nonlinear model is:
(4) The nonlinear mapping model is updated online, errors of real values and predicted values of real-time data fragments are calculated, the nonlinear mapping model is updated quantitatively according to the errors, output weights of the nonlinear mapping model are calculated in a resolving mode, the predicted value of the optical thickness of the aerosol of the next data fragment is obtained, online prediction of the optical thickness of the aerosol is achieved, and multisource features are fused through feature extraction, the nonlinear model and model updating strategies, so that online prediction of the optical thickness of the aerosol is achieved;
the nonlinear mapping model updating method comprises the following steps:
in the initial stage, output weights are obtained:new sample fragment->After arrival, the errors of the true value and the predicted value of the real-time arrival data fragment are calculated and expressed as:
wherein ,N0 and N1 Respectively representing the starting position and the stopping position of the current sample fragment, H 1 and y1 Hidden layer output and tag set, beta, representing newly arrived segments, respectively 1 Representing the current nonlinear model output weight, and obtaining beta based on the current output weight 1 The expression is:
considering the above expression in two parts, an intermediate variable is introducedAccording to the matrix multiplication combining law, the expression of the former part of matrix is:
in the current fragment state, H 1 and K0 Are known conditions and are further substituted into the integral output layer weight expression, so that the output weight beta of the model updating stage can be obtained 1 Is represented by the expression:
further performing iterative calculation to obtain an intermediate variable K corresponding to the (k+1) th segment k+1 Expression type
Further performing iterative computation to obtain the nonlinear mapping model output weight beta corresponding to the (k+1) th segment k+1 Is the optimal solution of (a):
the model expression after online updating is as follows:
substituting the input parameter x into the updated model, and calculating to obtain the aerosol optical thickness y, so as to realize online prediction of the aerosol optical thickness.
2. The online aerosol optical thickness prediction method based on multi-source information according to claim 1, wherein the multi-source information acquisition system in the step (1) comprises a data acquisition device, an information interaction terminal and a database, the data acquisition device comprises a laser radar, a ground weather station, a background radiometer and a corresponding data interface, the database stores data by adopting an SQL Server, the information interaction terminal performs information interaction with the data acquisition device, and acquired data information is stored in the database; the ground weather station, the background radiometer and the laser radar are used for realizing acquisition of atmospheric environmental parameters and optical parameters, wherein the atmospheric environmental parameters comprise atmospheric data wind speed WS, temperature TE, humidity HU provided by the ground weather station and sky background brightness LU acquired by the background radiometer, and the optical parameters are aerosol optical thickness and are acquired by the laser radar.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110898847.6A CN113657028B (en) | 2021-08-05 | 2021-08-05 | Online aerosol optical thickness prediction method based on multi-source information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110898847.6A CN113657028B (en) | 2021-08-05 | 2021-08-05 | Online aerosol optical thickness prediction method based on multi-source information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113657028A CN113657028A (en) | 2021-11-16 |
| CN113657028B true CN113657028B (en) | 2023-10-20 |
Family
ID=78490420
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110898847.6A Active CN113657028B (en) | 2021-08-05 | 2021-08-05 | Online aerosol optical thickness prediction method based on multi-source information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113657028B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114279915B (en) * | 2021-12-24 | 2024-08-27 | 青岛镭测创芯科技有限公司 | Atmospheric particulate concentration inversion method and related components |
| CN115081557A (en) * | 2022-08-22 | 2022-09-20 | 北华航天工业学院 | Night aerosol optical thickness estimation method and system based on ground monitoring data |
| CN116843456B (en) * | 2023-08-29 | 2023-11-07 | 北京燕知信科技服务有限公司 | Financial big data processing method and system based on artificial intelligence |
| CN117350440B (en) * | 2023-12-04 | 2024-07-12 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Space-time prediction model and method for optical thickness of regional aerosol |
| CN118211052A (en) * | 2024-04-11 | 2024-06-18 | 中国华云气象科技集团有限公司 | Multi-parameter meteorological data analysis and prediction method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106933585A (en) * | 2017-03-07 | 2017-07-07 | 吉林大学 | A kind of self-adapting multi-channel interface system of selection under distributed cloud environment |
| CN111861784A (en) * | 2020-06-01 | 2020-10-30 | 大唐东北电力试验研究院有限公司 | Photovoltaic power generation short-term prediction method based on artificial bee colony optimization neural network |
| CN111859800A (en) * | 2020-07-15 | 2020-10-30 | 河海大学 | Method for spatio-temporal estimation and prediction of PM2.5 concentration distribution |
-
2021
- 2021-08-05 CN CN202110898847.6A patent/CN113657028B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106933585A (en) * | 2017-03-07 | 2017-07-07 | 吉林大学 | A kind of self-adapting multi-channel interface system of selection under distributed cloud environment |
| CN111861784A (en) * | 2020-06-01 | 2020-10-30 | 大唐东北电力试验研究院有限公司 | Photovoltaic power generation short-term prediction method based on artificial bee colony optimization neural network |
| CN111859800A (en) * | 2020-07-15 | 2020-10-30 | 河海大学 | Method for spatio-temporal estimation and prediction of PM2.5 concentration distribution |
Non-Patent Citations (2)
| Title |
|---|
| Application of an improved artificial bee colony algorithm to inverse problem of aerosol optical constants from spectral measurement data;Zhenzong He et al.,;Optik;316-329 * |
| 无线光通信中的大气影响机理及抑制技术研究;陈纯毅;中国博士学位论文全文数据库 信息科技辑;I136-42 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113657028A (en) | 2021-11-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113657028B (en) | Online aerosol optical thickness prediction method based on multi-source information | |
| CN109214592B (en) | Multi-model-fused deep learning air quality prediction method | |
| CN109523013B (en) | Air particulate matter pollution degree estimation method based on shallow convolutional neural network | |
| CN115347571B (en) | Photovoltaic power generation short-term prediction method and device based on transfer learning | |
| CN111815033A (en) | Offshore wind power prediction method based on RCNN and meteorological time sequence characteristics | |
| CN111310965A (en) | Aircraft track prediction method based on LSTM network | |
| CN111461413B (en) | Highway road surface performance detecting system | |
| CN110751318A (en) | IPSO-LSTM-based ultra-short-term power load prediction method | |
| CN112070272B (en) | Icing thickness prediction method and device for power transmission line | |
| CN116957838A (en) | Crop growth environment monitoring method based on knowledge graph representation learning | |
| CN114444378A (en) | Short-term power prediction method for regional wind power cluster | |
| CN107703554A (en) | The warm and humid profile Inversion System of multichannel millimeter wave radiometer and its inversion method | |
| CN112818608A (en) | Medium-and-long-term runoff forecasting method based on improved particle swarm optimization algorithm and support vector machine | |
| CN109214591B (en) | Method and system for predicting aboveground biomass of woody plant | |
| CN114399073A (en) | Ocean surface temperature field prediction method based on deep learning | |
| CN115204032A (en) | ENSO prediction method and device based on multi-channel intelligent model | |
| CN114154401A (en) | Soil erosion modulus calculation method and system based on machine learning and observation data | |
| CN116244647A (en) | Unmanned aerial vehicle cluster running state estimation method | |
| CN116992779A (en) | Simulation method and system of photovoltaic energy storage system based on digital twin model | |
| CN113095547B (en) | Short-term wind power prediction method based on GRA-LSTM-ICE model | |
| CN117669008A (en) | Foundation settlement prediction method and system based on deep learning | |
| CN114925947B (en) | Phenological adaptive crop physiological index deep learning estimation method and system | |
| CN116189794A (en) | Rammed earth water salt content measurement method | |
| CN115907204A (en) | Forest transpiration water consumption prediction method for optimizing BP neural network by sparrow search algorithm | |
| CN115630566A (en) | Data assimilation method and system based on deep learning and dynamic constraint |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |